Bearing fault diagnosis based on multi-scale mean permutation entropy and parametric optimization SVM

Here, aiming at problems of difficult feature extraction and low accuracy of pattern recognition in rolling bearing fault diagnosis, a fault diagnosis method based on multi-scale mean permutation entropy (MMPE) and grey wolf optimized support vector machine (GWO-SVM) was proposed. Firstly, MMPE was used to comprehensively characterize rolling bearing fault feature information. Then, features with appropriate dimensions were selected to form a sample data set. Finally, GWO-SVM classifier was employed to do fault pattern recognition. The proposed fault diagnosis method based on MMPE and GWO-SVM was theoretically analyzed and studied, the corresponding test analyses were contrastively performed by using test data of rolling bearing. The results showed that MMPE can effectively extract the fault feature information of rolling bearing; the recognition accuracy and recognition speed of GWO-SVM are better than those of other commonly used parametric optimization SVMs of rolling bearing fault diagnosis; the proposed method can effectively identify fault position and fault degree of rolling bearing, the fault recognition accuracy based on MMPE and GWO-SVM is 98.0% using rolling bearing data set, it is higher than that based on MPE and GWO-SVM of 97.0%; the recognition accuracy based on MMPE and GWO-SVM is 93.5% under noise background, while that based on MPE and GWO-SVM is only 83.0%, so the proposed MMPE has better noise robustness. © 2022, Editorial Office of Journal of Vibration and Shock. All right reserved.